The question is no longer whether AI tools will enter the medical English writing classroom. They already have. The more pressing question is how to integrate them purposefully, without sacrificing the development of genuine language skills or undermining academic integrity. A wave of recent research across medical English, English for Academic Purposes (EAP), and English for Specific Purposes (ESP) is beginning to supply some answers.
The medical English case has been building for some time. A 2024 study published in BMC Medical Education examining AI-assisted academic writing among non-native speaking medical students found that students who received AI-assisted instruction showed improved proficiency across multiple aspects of writing, including organisation, coherence, grammar, and vocabulary (Li et al., 2024).¹ Encouragingly for hard-pressed educators managing large cohorts, statistical analysis revealed no significant difference between manual scoring and GPT-4 scoring, pointing to the potential of large language models to assist teachers in the grading process (Li et al., 2024).¹
The English for Academic Purposes (EAP) literature now adds considerable weight to these findings. A 2026 study published in Discover Education analysed postgraduate non-native English speakers enrolled in a mandatory EAP writing course. It found that generative AI-assisted writing feedback, when used with pedagogical awareness, can effectively enhance EFL learners’ writing skills and support active participation in the writing process. However, the researchers also stressed that critical thinking skills are essential and that students should avoid over-reliance on AI tools (Zhang et al., 2026).² Meanwhile, a 2025 article in ELT Journal exploring structured AI chatbot use in an online EAP module reported that unstructured AI use had led to increased student anxiety and poor learning outcomes, while structured use helped address those challenges. This is a timely reminder that education design matters as much as the tools themselves (Webb & Şenaydın, 2025).³
One of the most practically useful threads in the recent literature concerns the nature and quality of AI-generated feedback. Research synthesised in a 2025 review published by De Gruyter found that teacher feedback tends to focus more on content, while GenAI feedback focuses more on organisation (Xu & Guo, 2025).⁴ This suggests the two work best in combination rather than in competition.
A 2026 study in Social Sciences and Humanities Open comparing types of AI prompts found that EFL-tailored prompts produced the highest quality ChatGPT feedback in terms of relevance, clarity, usefulness, and tone (Algobaei & Alzain, 2026).⁵ The implication for medical English teachers is direct: prompts specifically oriented to clinical register, case report conventions, or patient communication genres are likely to produce more useful feedback than generic grammar-checking instructions.
For institutions uncertain about where to draw the line between legitimate scaffolding and problematic dependence, a framework published in TESOL Journal in early 2026 offers a practical starting point. The EAP AI Assessment Scale (EAP-AIAS) consists of five levels ranging from ‘No AI’ to ‘Full AI’. Each delineates appropriate GenAI usage in EAP tasks, and frames GenAI as mediation that can extend learner performance beyond what is reliably achievable independently (Tregubova, 2026).⁶ The scale reflects growing recognition that existing EAP curricula may inadequately prepare students to incorporate GenAI tools appropriately into their academic writing practices (Wu et al., 2026).⁷
A parallel framework, published in Artificial Intelligence in Education in 2026, applies the AI Assessment Scale specifically to EFL writing and translation contexts, demonstrating how different levels of AI assistance can be mapped to different stages of task completion and learner proficiency (Roe, Perkins, & Furze, 2026).⁸ For medical English programmes, either framework could be adapted to map AI use against the different genres students need to master, such as clinical case reports, patient discharge summaries, research abstracts, and reflective portfolios – with different levels of AI assistance deemed appropriate at different stages of proficiency and task complexity.
The picture that emerges across all of this research is nuanced but actionable. AI works best as a scaffolding tool — helping students draft, revise, and self-edit — while the teacher’s role shifts towards higher-order feedback on register, precision, and clinical appropriateness.
A 2025 study in Frontiers in Artificial Intelligence reviewing external variables influencing students’ attitudes towards AI writing tools found that motivation, engagement, and positive societal expectations are significant factors in determining whether students use these tools effectively – and that educators play a key role in shaping all three (Mansoor, Sumardjoko, & Sutopo, 2026).⁹
For medical English educators, the opportunity is not just to adopt these tools but to shape how they are used. The domain-specific knowledge that EMP specialists bring – understanding of clinical genres, register variation, the communicative demands of different healthcare settings – is precisely what is needed to ensure that AI writing support in medical education serves real language development goals, rather than simply producing polished text that bypasses them.
Back to Menu ↩ A concept whose time has come There is a phrase that has been quietly reshaping workplace learning and development over the
Back to Menu ↩ Beyond the novelty argument Mention gamification in a staffroom and reactions tend to split along familiar lines: enthusiasts who have seen
Back to Menu ↩ A new kind of practice partner There is a familiar tension at the heart of clinical communication training: students need to
